separation, what remained were diagnostic patterns of black spots on an X-

ray film. Each spot corresponds to an RNA fragment only five or six
nucleotides long. With effort, the exact nucleotide sequence of each
fragment could be ‘puzzled out’ by considering its migration during
electrophoresis and the enzymes that gave rise to it. In Woese’s day, each
RNA sequence became an entry in the lab RNA catalogue, analyses of
which eventually led to the discovery of the third domain of life, the
Archaea.

DNA sequencing
RNA-based techniques for molecular sequencing are laborious in the
extreme, and in the late 1970s they were quickly supplanted by two DNA
sequencing methods developed independently by Sanger and the American
molecular biologist Walter Gilbert. Gilbert’s ‘chemical sequencing’
approach is still useful in certain research applications, but here we will
focus on what became known as Sanger sequencing, as this is the one that,
for various reasons, took hold within the research community. It is still
widely used today, albeit with modern instruments and reagents.

Sanger’s method typically starts with the DNA fragment to be sequenced

inserted into a circular plasmid molecule to which a short DNA sequencing
‘primer’ has been bound. These DNAs are then mixed with a DNA
polymerase enzyme, the same enzyme that in nature replicates DNA in the
cell. To this mixture, each of the four building blocks of DNA is added:
these are the deoxynucleotides dATP, dCTP, dGTP, and dTTP, one of which
is tagged with radioactivity. Critical to the success of the experiment is the
addition of dideoxynucleotides (ddNTPs) as well. These ddNTPs act as
chain terminators: they are chemically modified in such a way that they can
be incorporated into DNA by the polymerase but cannot be used for further
chain extension. Classical Sanger sequencing uses four separate reaction
vessels for each DNA sample to be sequenced, which differ only in the
identity of the ddNTP added (each tube contains a different DNA chain
terminator).
In developing his procedure, Sanger’s brilliance was to tweak the relative
amounts of dNTPs and ddNTPs present in the reaction tubes such that
ddNTP incorporation, and thus chain termination, happens only rarely. In
the A tube, each time a T residue is encountered on the DNA template the
polymerase enzyme will add a normal dATP most of the time. This is
because A pairs with T and the tube contains more dATP molecules than
ddATPs. Chain elongation can continue. Upon reaching the next T in the
template, chain termination is again unlikely, but possible. The end result is
an A tube full of radioactive DNA fragments of multiple sizes, each
indicating the position of a T residue in the DNA template. The reactions
taking place in the three additional tubes are identical to the first, except
that the newly synthesized DNA fragments are terminated by ddCTP,
ddGTP, and ddTTP.

The final step is to use gel electrophoresis to separate the DNA fragments in
each of the four tubes side by side. This results in a DNA sequencing
‘ladder’, which (because of the use of radioactive dNTPs) can be visualized
by exposing a piece of X-ray film to the gel. By analysing the migration
patterns of the rungs on the ladder, the sequence of the original DNA
molecule can be inferred.

Sanger’s original version of dideoxy chain-termination sequencing was a

quantum leap beyond RNA cataloguing. Because it involved DNA
synthesis in a test tube, it sidestepped the need to obtain large quantities of
starting material (limiting which nucleic acids could be sequenced). It was
nevertheless still time-consuming and subject to human error (I speak from
experience). The gels needed for DNA separation are messy to prepare, and
the nucleotide sequences are read off the X-ray films by eye. Typically only
150–200 bp of sequence can be read per reaction, meaning that for longer
DNA fragments, multiple-sequencing reactions using different DNA
primers need to be performed in an iterative fashion.

In response to ever-increasing demand, Sanger sequencing has been refined

considerably over the years. In the mid-1980s, experimentation with the use
of fluorescent (as opposed to radioactive) labels led to the development of
‘automated’ DNA sequencers. With fluorescence, chemically distinct tags
can be used for each of the four ddNTPs, meaning that all possible chain
termination products can be synthesized in the same tube and run in the
same lane of a sequencing gel (see Figure 3). The use of heat-stable DNA
polymerases, like those used in the ‘polymerase chain reaction’ (a technique
for amplifying defined stretches of DNA from minute quantities of starting
material; see Chapter 5), helped to further automate and speed up the
process. Modern Sanger sequencing instruments separate the labelled DNA
fragments by passing them through fine capillary tubes; the machine (not a
human being) reads and records the DNA sequence by shining a laser on
the sample as it runs through the bottom of the tube and detecting which of
the four fluorescent tags are present.

3. Di-deoxy chain terminator (Sanger) sequencing (automated procedure).

DNA sequencing technologies have taken several more quantum leaps over
the past decade. Before moving on, however, it is worth noting that there is
still a niche for Sanger’s method. Using current instrumentation, it is still
the gold standard in terms of data quality, and it produces very respectable
read lengths (800–1,000 bp on average). Automated Sanger sequencing is
also the most cost-effective method if one’s needs are modest (e.g. if fewer
than several dozen short DNA molecules need to be sequenced). Note too
that we will forgo discussion of genome assembly until Chapter 3, after
exploring the various ‘next-generation’ sequencing technologies. By and
large, the challenges associated with stitching a genome together from
individual sequencing reads are the same, regardless of how those reads are
generated.

Next-generation DNA sequencing

At its peak, the government-funded Human Genome Project relied on the
sequencing capacities of more than a dozen universities and research
centres in the US, the UK, France, Germany, Japan, and China. Each was
filled with dozens of state-of-the-art Sanger sequencing instruments, each
sequencing 96 or 384 DNA samples in parallel, twenty-four hours a day,
seven days a week. Data quality was high and the task of sequencing the
human genome was completed ahead of schedule. But it was expensive—
US$3 billion expensive. The privately funded project, carried out by
American Craig Venter’s Celera Corporation, began much later and
produced a genome sequence of roughly the same quality with the same
instrumentation for ~US$300 million. (In doing so, Celera took advantage
of years of preliminary data generated by the public project; see Chapter 3
for discussion of the different strategies employed.) Having been
continuously tweaked and streamlined, Sanger sequencing had by the late
1990s approached the limit of data generation per instrument per unit time.

What emerged in the mid-2000s, shortly after the public and private human
genome projects were declared finished, was a sequencing technique
capable of generating orders of magnitude more data at a fraction of the
cost. The approach was called ‘pyrosequencing’, and what it shares with all
of the so-called ‘next-generation’ technologies that have come after it is